Define whether pause stops entries or position management

A pause command needs to name the activity it stops. We would distinguish a pause on new exposure from a stop on agent decision making and from an instruction to close positions. These produce different behavior when an account already holds risk.

An illustrative account holds a $5,000 long with an existing protective order. The owner chooses “pause new entries.” Under the proposed contract, opening a new position and increasing the long are blocked. Reading state and proposing a verified reduction remain permitted. Existing venue orders continue according to their own state unless the owner separately authorizes cancellation.

Use a table rather than one status bit

The pause record can identify permissions for observation, new entry, increase and reduction. It should also describe how previously active execution mechanisms are treated. A single boolean called paused cannot carry all of those meanings unless the surrounding contract fixes a single documented interpretation.

We would make release explicit. A model saying that conditions have improved supplies an observation, while the owner’s authorized resume action changes the paused permission. If the user specified an automatic resume condition, compile its exact predicate and data dependencies instead of allowing the agent to invent one later.

Our proposed replay begins with the held long, applies the entry pause and presents three actions: add $1,000, reduce $1,000 and reverse into a short. The first is rejected. The reduction may pass after position verification. The reversal contains new exposure and requires the same entry restriction as any other fresh position.

A second case pauses all agent turns while leaving the venue’s existing protective order active. Record that distinction plainly: lack of new model decisions does not cancel an already accepted venue instruction. A third case chooses closure, with its own authority and execution procedure. The three outcomes demonstrate why the terms need separate controls.

The review artifact should show current exposure and outstanding protection beside the requested pause scope. The owner can then understand what continues after the command. We would judge the implementation by permitted-action fidelity and trace completeness, with any market outcome remaining a separate observation.

Our mandate compiler provides the wider instruction-to-action framework for this pause permission table. The published controls research supplies its historical background. DXAP publicly describes the corresponding separation between model proposals and external policy checks.